FUNCTIONAL BRAIN MAPPING AT 9.4T USING A NEW MRI COMPATIBLE ELECTRODE CHRONICALLY IMPLANTED IN RATS

Abstract

There is a need for acute and chronic stimulation of the brain within the MRI for studies of epilepsy, as well as of deep brain stimulation for movement and behavioral disorders. This paper describes the production and characteristics of carbon fiber based electrodes for acute and chronic stimulation in the brain. Increasing MRI field strengths are making it increasingly difficult to introduce foreign objects without a susceptibility artifact. This paper describes the production of, and the characteristics of carbon fiber based electrodes. These are biocompatible and can be implanted for chronic studies. We show the use of these electrodes at 9.4T for studying functional activation. Data are presented showing regional connectivity. Activation not only occurs near the electrode, but at sites distant and often contralateral to the electrode. In addition, there were sites showing strong negative activation to stimulation both with direct stimulation and during a kindling associated seizure.

INTRODUCTION

Chronic brain stimulation is used to treat people with various neurological disorders including movement etiologies, epilepsy and a variety of other behavioral anomalies. The use of chronic stimulation in animal models provides us with valuable insights into brain mechanisms, function and connectivity patterns that underlie such disorders (eg. (1-3)). Magnetic resonance imaging (MRI) can be applied to study regional connectivity in brain with localized stimulation. It can also be used to study structure, physiology and connectivity during the initiation and propagation of seizure activity (4), but only if the electrodes are compatible for use at the high field strengths of the MRI system.

Magnetic field strengths for MRI systems for humans have reached 9.4T and for animals have reached 17.6T. Although these field strengths provide additional signal/noise, they also present a more difficult environment for using conductive materials. Many materials that conduct current will induce a very large susceptibility artifact in the MR system. Some metals and alloys can be used at field strengths up to 4.7T, (5), but it is more likely that carbon fiber based electrodes will have compatibility with the very high fields that are being implemented.

The goal of the current project was to develop electrodes compatible with chronic implantation and a 9.4T MRI and to apply them to observe functional connectivity in brain. We visualized connectivity using fMRI which is sensitive to the changes in deoxyhemoglobin associated with neurovascular coupling (6).

MATERIALS AND METHODS

Electrode production

Electrodes were constructed from bundles of individual carbon fibers (Formosa plastics corporation, Taipei, Taiwan) to achieve a final thickness of 0.2mm to 0.4mm diameter and a length of 10mm to 20mm. Other carbon fibres from Minnesota wire, as well as gold and platinum wires, were rejected as they showed significant susceptibility artifacts in the 9.4T MRI. The carbon fibers were attached to the slot of a 80 × 1/8″ brass screw (Spaenaur, Kitchener, ON) using silver print #842 (MG Chemicals, Toronto, ON) as a conducting adhesive. Carbon fiber bundles were insulated using the nonconductive polymer polyvinylidene difluoride (PVDF) (Whitford Corporation, Frazer, PA). PVDF is a thermoplastic that melts at high temperatures and hardens when cooled. PVDF was diluted with methyl isolbutyl ketone to achieve a viscosity that permitted individual carbon fibers to be lowered into solution without splaying. Electrodes were coated by dipping each electrode into the solution over 30s, using a clamp attached to a motorized pulley that controlled the speed. Electrodes were baked at 200°C for 10-15 minutes and cooled for 20 minutes between coats. Dipping was repeated three times to ensure sufficient insulation from electrode tip to the head of the brass screw. Before implantation, electrodes were cut to achieve the desired length and to remove insulation from the tip.

Animals

Ten male Long-Evans hooded rats weighing 323-356 g at the time of electrode implantation surgery were used in this study. Rats were obtained from the University of Calgary Breeding Colonies. They were housed individually in clear plastic cages with an enrichment toy in a colony room that was maintained on a 12 h on/12 h off light cycle, with lights on at 7:00 am. All experiments were conducted in the light phase. Rats received Lab Diet #5001 (PMI Nutrition International, Brentwood, MO) and water ad libitum. The rats were housed and handled according to the Canadian Council on Animal Care guidelines.

Electrode Implantation

Rats were anesthetized with an intramuscular injection of 58.83 mg/kg ketamine (85%) and xylazine (15%) at 1.0 ml/kg. Supplemental injections (0.05 ml) were administered as required. Lidocaine (2%) was administered subcutaneously at the incision site. Electrode and screw holes were drilled into the skull. One or two electrodes were then chronically implanted in the right hemisphere of the sensorimotor neocortex according to stereotaxic coordinates (7). One electrode was implanted in the callosal white matter (1.0 mm anterior to bregma and 0.5 mm lateral to midline) and in most animals a second electrode was implanted in the right frontal neocortex (1.0 mm anterior to bregma and 4.0 mm lateral to the midline). Electrodes protruded 4.0 mm from the skull surface to increase the distance of the brass screw from the brain. Four plastic screws 8L080X093N01 (Plastics One, Roanoke VA) were screwed into the skull surrounding the electrodes. Electrodes and plastic screws were bound together with dental cement. Experimental procedures commenced no earlier than 7 days after surgery.

Electrical stimulation to elicit seizures

The rats were divided into two treatment groups; sham-stimulated and stimulated. On day 1 of stimulation, an afterdischarge (AD) threshold was determined for each rat in the stimulated group. AD thresholds were defined as the weakest current necessary to induce an AD (also known as epileptiform activity) of at least 4 seconds. Current was delivered through both implanted electrodes. The stimulation consisted of a 1 s train of 60 Hz biphasic rectangular wave pulses, 1 ms in duration and separated by 1 ms. The current commenced at 50 μA, and was increased in steps of 50 μA every 60 seconds until an AD of at least 4 s was recorded. Thereafter, stimulation was delivered once daily at an intensity 100 μA greater than the initial AD threshold. Sham-stimulated rats were treated in the same manner, except that they did not receive any current. Seizure behaviours were scored according to a five-stage scale (8).

EEG Recording

Neocortical EEG recordings were digitized from the monopolar electrode using a Datawave 12-bit analogue to digital (A/D) board for up to 6 weeks after implantation. The callosal electrode served as a ground reference during the recording. Signals were sampled at 500 Hz using Grass Model 12 EEG amplifiers and filtered at half amplitude below 0.3 Hz and above 300 Hz. Signal analysis was performed using SciWorks Data Acquisition/Analysis package (Datawave, Longmont, CO). A Fast Fourier transform (FFT) was performed on a five second stable segment of EEG recording for each rat.

Rat stimulation in the MRI

Rats were anesthetized with isoflurane (1.5-2.5% in 30% oxygen, remainder nitrogen) and placed within an MR cradle. The head was restrained using ear pins and a bite bar and a surface coil was positioned over the cerebrum. Animals were allowed to breathe spontaneously and respiration and temperature were monitored in the magnet. Rectal temperature was measured and core body temperature maintained (37.0-37.5 °C) using a feedback heated air system.

Stimulation was delivered using a Grass stimulator with a stimulus isolation unit which applied a 1 mA constant current with a stimulus train of 2ms square wave, 60Hz for 2s. Varying numbers of stimulation trains were repeated from single train to a set of 6 stimulation trains.

MR Imaging

MRI was done with a 9.4T Bruker Avance system, a birdcage coil (for ex vivo electrode imagingn) and 3 cm diameter surface coil for in vivo MRI. MRI scans of electrodes and parts of the electrode systems were first acquired to assess their potential to induce susceptibility artifacts in the images. For these studies, both gradient and spin echo images were obtained through electrode components that were either embedded in agar, suspended in saline or imaged in vivo. Imaging for electrode susceptibility was done using a gradient echo FLASH sequence with (TR/TE/α = 100 or 61 ms/ 4.5 or 5 ms/ 20°).

Functional magnetic resonance imaging (fMRI) of anesthetized animals was done with a fast spin echo sequence (TR/TE = 2.26 sec/58.9 ms, 64 echoes per TR, FOV=3 cm², matrix size= 128×128, slice thickness 1.5 mm). Shimming was done on a slab including the region of interest and signal magnitude was optimized with first and second order shims. Each fMRI series consisted of 10 baseline images prior to stimulation and then images collected during and following a series of stimulations with a minimum of 10 images between stimulations to follow recovery. The sets of images were analyzed for regions of activation using a fuzzy-cluster based analysis software (EvIdent™, Inst for Biodiagnostics, NRC). This detects clusters of voxels with self-same intensity changes (correlation to average time course with p<0.0005) of positive or negative BOLD signal changes (9).

Histology

A subset of rats were anaesthetized with an overdose of sodium pentobarbital (100mg/kg). The brains were extracted, frozen on ice, and stored at -80°C. The brains were sectioned in the coronal plane at a thickness of 30 μm using a cryostat.

RESULTS

Electrode Characteristics

The electrode was comprised of a bundle of carbon fibers connected to a brass screw with silver print and coated with PVDF (Fig. 1a). Electrodes were tested for resistance, impedance, and ability to deliver square wave pulses. The resistance of the electrodes ranged from 5 to 20Ω, the impedance was below the kilo ohm range, and no attenuation of the square wave pulses was observed.

Figure 1.

a: photograph of electrode with a cm scale bar. b: coronal GE FLASH image of a brain slice containing the implanted electrode near corpus callosum showing minimal susceptibility artifact c: sagittal image of same electrode (TR/TE/α = 100 ms/ 4.5 ms/ 20°).

Numerous coatings were investigated with variable results. We found methylmethacrylate to be brittle and easy to crack. This was the same problem with commercial nail polish (which also had the potential of uncertain biocompatibility). Teflon and polyamides were tried, but found to be more difficult to handle. PVDF (or Kynar) was found to be suitable with acceptable biocompatibility (Fig. 2). PVDF is resistant to radiation, chemical degradation and heat, which provides for a range of methods for sterilization before implantation. It is also tough and slightly flexible, making the electrode very robust, and strong enough to push through most tissue.

Figure 2.

Histological section of rat brain from region of electrode implant 8 weeks post implant. a,b: Low and high magnifications, respectively, of representative Nissl-stained coronal sections through the track of a stimulating carbon electrode. The arrow indicates the region near the tip. Although the track shows structural damage tissue surrounding the tip appears normal. A. X2.5; B. X10. c: Electrical recordings of chronically implanted metal wire electrodes in rat brain. This is the standard, non-magnetic compatible electrode used previously for rat studies d: similar recordings with the carbon fiber electrode. The top trace is a recording of spontaneous activity under isoflurane anesthesia. The bottom trace is the fourier transform, showing the frequency patterns in the rat cortex. The pattern of recording is similar with the two electrodes. The mv response is the same or higher for the carbon fibers, indicating that they have appropriate conductive properties for both stimulation and recording.

PVDF, in combination with the carbon fibers used here, had minimal susceptibility artifact as seen in a gradient echo image Fig 1b. It is important to note that not all carbon fibers were compatible with high field MRI. We investigated 3 cables from Minnesota Wire & Cable Company--products 12702, 18073, and 18078. Each of these had significant susceptibility artifacts. The electrode itself in Fig. 1 is approximately 400μm in diameter. The electrode was approximately 2-3 pixels across in the GE image (voxel size of 234μm) and so the susceptibility artifact was, at maximum, about 300μm. The electrode could not be seen in the spin echo images.

EEG and Stimulation in non-MRI environment

Robust and artifact free EEG, similar to that recorded from stainless steel electrodes was recorded from unanaesthetized, freely behaving rats (Fig 2). Both carbon and steel electrodes recorded low frequencies 1-20 Hz and some higher frequencies 20-50 effectively. Although there were individual variations in power, these may relate to placement. On the whole, the power at each frequency was comparable between the carbon fiber and stainless steel electrodes. The impedances between the two types of electrodes were not different and actually overlapped in measurement values. There were no discernable differences between the electrodes over time and between rats. Stimulation through the carbon fiber electrodes yielded epileptiform activity and seizure behaviour that was also comparable to stimulation through stainless steel electrodes (1).

Stimulation in MRI environment

Direct electrical stimulation of the brain generated both positive and negative BOLD responses as evidenced by cluster analysis. Areas near the electrode showed a positive BOLD response (Figs 3-4). Patterns of activation varied. Trains of 1-6 repeated stimuli were used. Significant BOLD effects were not always observed with 1 train, but were always observed with higher numbers of train repetitions. Higher repetitions resulted in strong activation on both the ipsilateral and contralateral sides of the brain relative to the electrode implant (Fig 3). When an electrode was placed near the corpus callosum, there was strong activation around the electrode with negative BOLD responses on the lateral cortex, extending into the striatum (Fig 3d).

Figure 3.

Example of fMRI responses. Top shows mean % BOLD signal change for the voxels within that cluster correlating to the mean time course for active voxels in that cluster (p<0.0005). a,b: show similar slices from two different subjects. There were 6 consecutive stimulation trains resulting in 6 peaks of activation or positive BOLD responses. These animals had 2 electrodes (stimulation and ground) approximately 2mm apart. The bottom shows the results from one cluster where the fast spin echo image is overlaid with the active voxels (orange-red) correlating to the mean time course for that cluster (p<0.0005). Arrows show the regions of the two electrodes. There was strong activation in both the ipsilateral and contralateral cortex c) Example of activation responses from an animal with a single electrode implanted near the corpus callosum. Arrow indicates the electrode site d: the same subject as in “c” showing negative activation in the cortex which extends into the striatum. There were 6 repeated trains of stimuli.

Figure 4.

Activation maps from a single animal showing a,b: strong activation near and lateral to the electrode, and, c: activation up to 3 mm posterior to the electrode. d: Repeated stimulation trains (over 2) resulted in some negative activation in the hippocampus (largely ipsilateral in this image).

Three to four 1.5 mm slices were collected during the imaging. In posterior slices, regions located more than 2mm posterior to the electrode also showed BOLD activation (Fig 4). Thus, activations were not limited to the slice containing the electrode.

Large baseline drifts were often observed (as in Fig 3a). When stimuli were repeated in the same animal the BOLD responses were always similar in pattern and location. There was less reproducibility between animals, possibly because different electrode positions were being investigated. Only one subject did not show negative BOLD responses. Regions in cortex that showed BOLD increases never showed decreases. Subcortical regions were more variable in that both increases and decreases were observed.

Rats that had experienced repeated seizure activity prior to imaging showed patterns of activation that extended beyond the direct time of the stimulation (Fig. 5). BOLD signal enhancement took approximately 15 scans (60-70s) to recover to baseline. This contrasts with the response duration of approximately 5-8 scans (23-37s) in experimentally naïve rats. Large areas of focal activation were not observed, rather the responses consisted of scattered focal responses. Interestingly, there was both negative and positive BOLD effects observed during the presumed seizure.

Figure 5.

Response patterns during a seizure. Descending rows are data from subsequent seizures in the same animal. a: negative responses, with a possible rebound or overshoot. b: positive responses shown in the ipsilateral hippocampus c: small regions in both ipsilateral and contralateral parietal cortex. The duration of seizures, based on BOLD MRI, was in the order of 40 images or 160s.

DISCUSSION

There are significant advantages to using a chronic implant for brain stimulation, even if acute studies of activation patterns are to be undertaken. The chronic implant allows for the tissue to heal around the electrode, and reduces the complications of motion in during the study. Figure 2 shows that tissue has surrounded the electrode and there is no evidence of residual blood or cystic structures. The tissue does show significant damage along the electrode tract, as is expected with an electrode implant but the damage is not inflammatory. The coatings are biocompatible as there is no evidence of gliosis or unusual tissue structure at the tip of the electrode where the stimulation and recordings occur. If one were to use an acute electrode implant, as has been done by Shyu et al. (10), then there are increased complications related to trying to keep the electrode from moving relative to the animal during the study. The tissue will have acute trauma including bleeding, blood brain barrier disruption and cellular damage.

The susceptibility artifact with the GE imaging (TE=3.5ms) was about 300μm, approximately equal to the diameter of the electrode. The combined size of the electrode and artifact in the image was still less than the size of electrodes used for intracranial EEG recordings or deep brain stimulations (approximately 1.1-1.8mm). Fast spin echo, not EPI was used in vivo for stimulation studies. FSE sequence is less sensitive to susceptibility artifacts, we can’t state that the electrodes were tested with EPI in vivo. However the relatively low artifact obtained with GE imaging could be used to argue that the electrodes would be acceptable with EPI.

The advantage of constructing the electrodes from carbon fibers is that batches of these fibers can be obtained with little or no susceptibility artifact. With care, the electrode could be made significantly thinner than that used here (approximately 400μm), but these thick electrodes are very simple to produce. The electrical properties are conducive to both recording EEG’s and to transmitting activation currents. A previous study reported results from 3 electrode types tested at 4.7T—carbon fiber, gold and platinum-iridium (5). In that study, they found the carbon fiber, and platinum-iridium electrodes to have significant susceptibility effects. With respect to the carbon fibers, this may have been due to the batch, since we found a range of susceptibility artifacts with the 4 different carbon fibers that we studied. Recording electrodes of platinum-iridium have been used at 4.7T in monkey brain (11), but it is possible that they would accept a given size of artifact in a large brain that would be unacceptable in a rodent brain. Silicon electrodes have also been used in MRI (12). This is an elegant method but is not as readily producible was not tested within an MRI.

The PVDF coating is also well suited for this application. It is often used for electrode coatings, it is flexible and tough, and it is resistant to radiation, many chemicals and temperature (making sterilization relatively simple). In addition it is highly biocompatible. PVDF has been used routinely to coat electrodes for chronic electrical stimulation and recording studies (13). A rabbit, which had an intracortical metallic electrode implant coated with PVFD, was studied for 5 years with no adverse health and with perfect recordings (Dr. B.H. Bland, pers. comm).

The functional response of the brain was assessed by measuring the BOLD response (14), where increases in signal intensity correlate with areas of increased brain function. Areas around the stimulation electrode always show positive activation. In areas in direct contact with the electrode, the activation often extends millimeters away both across the brain (Fig. 3) and in the posterior direction (Fig 4).

The observation that stimulation was observed in regions distant from the electrode supports work showing that direct stimulation in brain causes activation that extends beyond that which can be attributed to direct electrical current propagation (4,15,16). This implies that the fMRI is detecting functional connectivity linked to the stimulation. This conclusion is also supported by the existence of activation clusters that are many millimeters from the electrode and across the midline, presumably indicative of transcallosal activation (Fig 3). Studies of changes in T2 in a chronic kindling model show morphological modifications distant from the electrode implantation site indicating that distant circuitry was activated (17). Changes in function of the contralateral auditory cortex have been observed after kindling, providing additional evidence that activation can spread across the midline (18)

The presence of negative BOLD responses highlights the complex interactions of brain to localized activation. In this example, the response patterns were both negative, and “narrower” than in the positive responses. It may be that these differing response patterns relate to the location of the responses in the vascular bed, a variable that has been noted to relate to response times (19). Although the peaks appear sharp and correlated to the stimulus we do not believe they are artifactual because the stimulus is 2s in duration whereas the intensity changes are approximately 18s in duration. Each image is approximately 3 sec in duration and a hemodynamic response to a short stimulus is expected to respond in 2-5 sec (20). The activation response in Fig 3D is about 6 images long with a maximum in the second image after stimulation onset, consistent with a tissue BOLD change and not directly related to the electrical stimulation.

The presence of negative BOLD responses (assuming these signals still represent neuro-vascular coupling) highlights the complex interactions of brain to localized activation. The response patterns may be “narrower” than in the positive responses but the limited data set precludes a strong conclusion. It may be that the differing response patterns relate to the location of the responses in the vascular bed, a variable that has been noted to relate to response times (19). In our case, activation of the corpus callosum and cortical regions near the corpus callosum resulted in negative cortical and striatal responses (Fig 3). The interaction between cortex and striatum is expected based on the feedback loops that are known to exist (21). Functional connectivity has also been shown in an fMRI study using rats and an acutely implanted electrode. In this study, stimulation of the thalamus was observed to have functional links to the anterior cingulate cortex (22).

We have produced electrodes which are compatible with high MR field strengths, which cause little or no biological reaction, and which can be used to either stimulate or record brain activity. We show a subset of data which indicate that there are complex response patterns elicited by stimulation from a chronically implanted electrode. There are interesting variations in the responses—both in terms of location and in the coupling between stimulus and response duration. Although these are not within the scope of this paper to study, it is clear that this chronically implanted preparation will provide novel information about brain connectivity as well as the changes in connectivity that may accompany chronic stimulation or the development of epilepsy. In future, we could also assess activation patterns in brains that have recovered from the trauma of surgery or stroke, and will allow us to monitor changes in activation patterns while the brain adapts to acute or chronic damage.